LoginSign Up
← Back to all sparks
R

remap

INFRA · APIS
Velocity0.0

Regional Spatial Modeling with Continuous Borders

github.com/jadonwagstaff/remap ↗

A regional-model smoother in long-term maintenance, four years past its last real feature.

spatial-modelinggeostatisticsr-packagesmaintenance-modesf-ecosystem
◆Current state
remap fits separate models to geographic regions and blends their predictions into a continuous surface, using the min_n nearest observations so region boundaries do not show as discontinuities. The methodological work is finished — the last behavioural change landed in 2022, and everything since has been unit handling, compiler warnings, dependency compatibility and citation updates. The July release is one such patch.
◆Where it's heading
This is a stable academic package tracking its ecosystem rather than growing. The visible pattern is reactive maintenance: sf's 1.0.0 transition, ggplot2's size-to-linewidth rename, a gcc-UBSAN error on zero-point distance calculations, and now a check that distance matrices passed to remap() and predict() are converted to kilometres. Note that the feed's timestamps invert the version order — 0.3.1 is stamped seconds after 0.3.2 despite being the earlier release, so recency in this feed is not a reliable guide to sequence.
◆Prediction
Nothing in these entries points to new capability. The realistic expectation is more of the same: a patch whenever sf, ggplot2 or a CRAN check surfaces an incompatibility, at roughly the observed cadence of one release every year or two.

◆Recent moves

  1. 1mo ago

    Distance matrices now converted to kilometres automatically

    A defensive patch: remap() and predict() check the units on a supplied distance matrix and convert to kilometres if needed, plus a website update for the utsnow dataset. Consistent with the package's post-2022 posture of correctness patches only.

    View source ↗
  2. 1y ago

    Citation updated to the R Journal article

    Citation pointed at the published R Journal paper and the suggested-package list trimmed. Nothing user-facing changes. Despite carrying the later timestamp in this feed, it precedes 0.3.2.

    View source ↗
  3. 1y ago

    UBSAN fix, parallel patch and sf search-path independence

    Ecosystem-tracking maintenance: a gcc-UBSAN error when computing distances for zero points, ggplot2's size-to-linewidth rename, a parallel computing patch, and package data restructured so it works without sf attached. The release notes state the reformatting carries no functional change.

    View source ↗
  4. 3y ago

    predict() can return an upper bound on combined standard errors

    The last release in this window to add capability: an se option on predict.remap that smoothly combines regional model standard errors into an upper bound, when the underlying models return them. Backwards compatible, and the end of the package's feature line as the feed records it.

    View source ↗
  5. 5y ago

    Regions without observations no longer break the prediction surface

    Removes the requirement that a region contain at least one observation before it gets a model; regions now draw on the min_n nearest observations regardless. This is what made small-region setups usable, since those were the cases where discontinuities appeared in the surface.

    View source ↗